In this article, the author provides an alternative to traditional portfolio re-balancing based on changes in asset market values, one informed by equity characteristics. The logic of policy portfolio re-balancing is applied to a framework that uses assets’ 12-month rolling average characteristic values and return volatilities as inputs and re-balances toward those assets that exhibit relatively high characteristic-value-to-volatility ratios. Three important practical constraints are applied to the author’s re-balanced portfolios. The first two constraints relate to asset weights and limit the degree to which a re-balanced portfolio’s individual asset class positions and portfolio-level asset allocation can deviate from a fixed-weight reference policy portfolio. The third constraint is that any re-balanced portfolio is required to have a volatility approximately equal to the representative fixed-weight policy portfolio. The author shows that the longer-term, full-sample performance, both risk and return, of even tightly constrained characteristic-driven policy portfolios is superior to the standard procedure of re-balancing to fixed weights. These results are further validated by measuring the performance of characteristic-driven policy portfolios over rolling sub-samples of the original dataset. In the final section of the article, a meta-optimization technique is introduced that allows investors to select one characteristic-driven policy portfolio among several as the best compromise asset allocation that comes closest to maximally satisfying each of their respective objectives. TOPICS:Portfolio theory, portfolio construction, equity portfolio management Key Findings • Policy portfolios based on re-balancing toward assets that exhibit relatively high characteristic-value-to-volatility ratios are shown to produce better performance relative to standard policy portfolios that re-balance towards fixed policy portfolio weights. • Characteristic-driven policy portfolios exhibit superior performance compared to standard policy portfolios even when the characteristic-driven policy portfolios are highly constrained with respect to their individual asset class positions, portfolio-level asset allocations and volatility profiles. • By means of a simple meta-optimization technique, it is possible to select one characteristic-driven policy portfolio among several as the one that comes closest to maximally satisfying each of their respective objectives.